Abstract
Imitation has long been recognized to serve as a learning tool that enables people to acquire new skills and knowledge and as a strategy to engage in social exchanges with others (Uzgiris, 1981). Imitation, because it is foundational to the learning and development of social communication skills (Uzgiris, 1981), is therefore targeted in many pediatric therapeutic interventions. Occupational therapists, for instance, often use task demonstration in their daily practice with children, with the aim of promoting cognitive development and acquisition of new motor skills (Liew et al., 2012).
Compared with their typically developing peers, children—and also adults—with autism spectrum disorder (ASD) exhibit significant deficits in imitation, particularly in the spontaneous use of imitation during social interactions (Edwards, 2014; Van Etten & Carver, 2015). Ingersoll (2008), for example, showed that children with ASD tend to perform worse on tasks assessing spontaneous imitation than on tasks measuring elicited imitation. Children with ASD also tend to perform less well on tasks involving the imitation of body movements and gestures that are arguably not meaningful than on tasks involving actions on objects or actions that have a clear meaning or visual goal (Zachor et al., 2010). Thus, it has been suggested that therapies designed to improve imitation in people with ASD, which are often used in occupational therapy, should adequately address the social use of imitation, notably by also focusing on increasing intrinsic motivation for social interactions (Van Etten & Carver, 2015). In other words, a therapist should focus not only on whether a child can learn a new skill through imitation but also on assessing and facilitating motivation to imitate and engage in social interactions through imitation.
Social motivation, as indexed by social orienting, social seeking and liking, and social maintaining, is often profoundly impaired in people with ASD (Chevallier et al., 2012). Both neurobiological evidence and behavioral data suggest that children with ASD tend to perceive higher social reward from contact with animals than from contact with people, and they show strong motivational preferences toward animals (Celani, 2002; Whyte et al., 2016). Also, data have been reported that suggest that interaction with animals may decrease social anxiety in children with ASD while also motivating or facilitating positive social behavior toward other persons, such as peers and therapists (O’Haire et al., 2015; Silva et al., 2011; for reviews, see O’Haire, 2013, 2017). Sams et al. (2006), for example, compared the performance of children with ASD receiving two forms of occupational therapy—occupational therapy using standard techniques and occupational therapy incorporating animals—and found that the children in sessions incorporating animals demonstrated significantly more use of language and significantly more social interaction than those in sessions using exclusively standard occupational therapy techniques. More important, no research on human–animal interactions to date has focused on adults with ASD (O’Haire, 2017).
On the basis of these findings, we aimed to provide a first, preliminary experiment of the potential benefits of dogs for children and adults with ASD, in the particular context of imitation. Specifically, we aimed to test whether promoting contact with a friendly dog before a spontaneous imitation task would affect participants’ performance of the task. A within-subject design was used to compare the effects of three experimental conditions that varied in whether the imitation task was preceded by a waiting period or by a free-play interaction with either a live dog or a robotic dog. Imitation ratio, imitation accuracy, and indicators of social motivation were assessed during the task. We tested whether, after free play with the live dog, participants would (1) engage in spontaneous imitation more frequently and more precisely than in the other two conditions and (2) appear more socially motivated than in the other two conditions. Also, we assessed whether greater engagement with the animal—but not with the robot—before the imitation task was associated with better performance of the task.
Method
Participants
Forty people were identified for study participation through two local agencies serving people with ASD. Inclusion criteria included a clinical diagnosis of severe ASD (established using the Autism Diagnostic Interview–Revised; Lord et al., 1994). Exclusion criteria included motor impairments, pet ownership, allergies to and fear of dogs, and abnormal sensory reactivity to the test stimuli. As a result of practical considerations (e.g., family agendas), family consent for study participation was obtained for 25 individuals (see Table 1 for participant characteristics). Participants’ symptom severity at the time of the study was assessed using the Autism Diagnostic Observation Schedule (Lord et al., 2000). All participants in this study had severe language impairments (they were either preverbal or used single words).
Participant Characteristics (N = 25)
Stimuli
The dog in this study was a 4-yr-old male Labrador Retriever certified as a therapy dog by the Associação Portuguesa para a Intervenção com Animais de Ajuda Social (a member of Assistance Dogs International). He was selected on the basis of previous involvement in dog-assisted intervention programs. The robotic dog that served as the live dog surrogate was the Zoomer dog from SpinMasterTM (Toronto). It can be programmed to act autonomously, emulating the behavior of a live dog (e.g., it can roll, sit, bark, and approach a person, similar to live dogs).
Experimental Procedure
Participants were tested in a quiet room, either at home (children) or at the day center in which they were enrolled (adults). A first meeting was arranged with each participant to obtain family written consent for study participation, familiarize the participant with the experimenters and the stimuli (the live dog and the robotic dog), and acquaint the animal with the settings. More important, during this first meeting no participant in the study showed abnormal sensory reactivity to the stimuli; all accepted touching and being touched by both the live and the robotic dogs.
The experimental procedure began within 48 hr after the first meeting. Two researchers (E1 and E2) were present during testing: One (E1) interacted with the participant, and the other (E2) video recorded the session and never interacted with the participant. As in O’Haire et al. (2015), the video camera was positioned on a tripod, approximately 10 ft (3.048 m) in front of E1 and the participant. It was monitored and adjusted by E2 to ensure that participants were in view at all times. The roles of E1 and E2 were kept constant throughout the study. In addition to the experimenters, and for ethical reasons, a person familiar to the participant was also present during testing. This person was instructed to refrain from engaging in any interaction with the participant during testing. The presence of a familiar person was aimed at avoiding additional stress to the participants.
All participants were involved in three test sessions—one test session per experimental condition—separated by a 1-wk washout period. Experimental conditions involved the live dog (dog condition), the robotic dog (robot condition), or no stimulus (no-stimulus condition). To achieve balance across participants, the order of conditions was determined through constrained randomization.
Depending on condition, sessions began with either a 5-min waiting period (no-stimulus condition) or a 5-min free-play interaction period with the stimulus (dog or robotic dog condition). This means that in the dog condition, for example, participants were presented with the dog and interacted with the animal before the spontaneous imitation task. The same procedure was followed in the robot condition. More important, E1 remained passive during this free-play period (pretending to be doing some paperwork) and did not attempt to influence the participant’s interaction with the animal or the robot. This was planned so that we could compare the participants’ engagement with each of the dogs (live and robotic). In the no-stimulus condition, the participant was told to wait a moment so that E1 could finish some paperwork before playing together. During this waiting period, the participants were free to act as they pleased.
The spontaneous imitation task began with E1 inviting the participant to sit on a chair opposite her so that they could play a game. In the live dog and robotic dog conditions, participants were told that the dog or robotic dog needed to rest but that they could play again together at the end of the session. To avoid inducing stress in the participants from the removal of the stimulus, the live dog and the robotic dog remained present during testing. The live dog was instructed to lay down and stay, and the robot was put in sleep mode.
The spontaneous imitation task was adapted from previous studies (e.g., Rogers et al., 2010). It involved a small battery that included four manual actions (clap hands, pat legs, touch nose, wave goodbye) and two orofacial actions (extend tongue, make a noisy kiss). We chose this small battery following clinical advice and in consideration of participants’ reduced tolerance to testing. Modeling of each action was paired with a verbal marker that included the participant’s name to draw attention to E1. Modeling occurred only when the participant looked directly at E1. No instruction for imitation was provided. Each action was modeled a total of three times, and the order of actions was randomized. Participants were given a response period of 5 s after each action was modeled.
This experimental protocol was approved by the ethics committee of the Institute of Biomedical Sciences Abel Salazar, Porto University (PROJ121/2015CETI).
Coding
Two independent raters coded the sessions using the Observer XT® software (Noldus Information Technology, Toronto). The participant’s response after each action was modeled was scored as in previous studies (Rogers et al., 2010): 0 (no movement), 1 (a contingent movement appearing unrelated to the modeled action), or 2 (some degree of imitation). Whenever imitation occurred, the number of errors that occurred in its production was coded according to the performance criteria described in Table 2 (and also following Rogers et al., 2010). Six categories of errors were considered: bilateral versus unilateral, position, location, dynamic, repetition, and direction.
Performance Criteria for Each Action Included in the Imitation Battery
Note. E1 = experimenter who interacted with the participant; NA = not applicable.
Two variables were generated: (1) imitation ratio (total number of “2” scores divided by the total number of imitation opportunities) and (2) imitation accuracy (mean number of errors in each response).
As indicators of social motivation (Chevallier et al., 2012), we assessed response to name (i.e., mean number of calls it took a participant to look at E1 before modeling) and participants’ emotional expressions during testing. Emotional expressions were coded in 5-s intervals ranging from 1 (laughing) to 4 (neutral) to 8 (cry). A mean emotional rating was calculated for each participant. Also, a coefficient of variation (CV) was calculated by dividing the standard deviation by the mean. The lower the CV, the more stable emotional expressions were and the less variation occurred in them.
As a measure of engagement with the stimuli, the total duration of the participants’ social contact (i.e., eye gazing and physical contact) with the dog and the robot was also coded. A preliminary viewing of the sessions revealed that participants rarely looked at or tried to touch the stimulus during the imitation task. We therefore coded only participant–stimulus interaction during free play.
Statistical Analysis
Data were tested for effects of group (children and adults) and condition (dog, robot, and no stimulus) using repeated-measures analyses of variance (ANOVAs). Spearman rank order correlations were done to examine whether greater engagement with the stimuli (dog and robot) before testing was associated with individual differences in imitation ratio and accuracy. Assumptions of statistical tests were checked as appropriate. Greenhouse–Geisser correction to degrees of freedom was applied when violations of sphericity were present (p < .05). Post hoc tests using Tukey’s Honestly Significant Difference method of contrasting individual treatments were carried out with Bonferroni correction. Interrater reliability was computed using Cohen’s κ for categorical variables and Pearson correlation coefficients for continuous variables. All analyses were done with IBM SPSS Statistics (Version 24; IBM Corporation, Armonk, NY), with significance set at p < .05.
Results
Table 3 shows Cohen’s κ, means, and standard deviations for all variables considered in this study. ANOVA results showed a significant Group × Condition interaction effect on imitation ratio, mean emotional ratings, and response to name (Table 4). Post hoc tests indicated significant differences between conditions only among the children. Imitation ratio was higher in the dog condition than in the other two conditions (dog vs. robot, p = .009; dog vs. no stimulus, p = .002; see Table 3). Emotional ratings were lower (thus, emotions were less negative) in the dog condition than in the other two conditions (dog vs. robot, p = .001; dog vs. no stimulus, p ≤ .001; see Table 3). Mean number of prompts before a response to name were also lower in the dog condition than in the other two conditions (dog vs. robot: p = .026; dog vs. no stimulus: p ≤ .001; Table 3). Post hoc tests also showed significant differences between groups. Children’s imitation ratio was higher than adults’ in the dog condition. Children’s emotional ratings were higher than adults’ in both the robot and the no-stimulus condition (p < .001 in both cases; Table 4). Mean number of prompts before a response to their name was also higher for children than for adults in these same conditions (p < .001 in both cases; Table 3). No significant effects of group or condition were found on the CV of the emotional ratings or on imitation accuracy (Table 4).
Descriptive Statistics for the Behavioral Variables
Note. N = 25, 15 adults and 10 children. CV = coefficient of variation; M = mean; NA = not applicable; SD = standard deviation.
p < .05.
Analysis of Variance of the Data
Note. N = 25 (15 adults and 10 children). CV = coefficient of variation; df = degrees of freedom; MS = mean squares; SS = sum of squares.
p < .05.
A significant Group × Stimulus interaction effect was also found for duration of social contact with the stimulus (see Table 4). Children engaged more with the dog than with the robot (p < .001; see Table 3), whereas the opposite was found for adults (p < .001; Table 3). In the dog condition, children engaged longer with the dog than did adults (p < .001; Table 3). Results of the Spearman correlations showed a marginally significant trend. In the dog condition, participants who spent more time engaging with the animal before testing tended to imitate more frequently (r = .390, p = .054).
Discussion
In this study, only the children showed differences in behavior across conditions that seem consistent with the possibility that free play with the live dog before the imitation task was effective in promoting spontaneous imitation. As opposed to adults, who showed no differences in imitation between conditions, children engaged more frequently in spontaneous imitation in the dog condition than in the other two conditions. Moreover, children showed more positive emotional expressions and needed fewer prompts before responding to their name in the dog condition than in the other conditions, thus suggesting higher levels of social motivation after free play with the live dog. No such difference between conditions was found in the adult group.
At this point, we can make no conclusions as to why the dog condition, compared with the other conditions, was associated with increased spontaneous imitation and seemingly higher social motivation only in the child group. When compared with adults, however, children engaged in more social contact with the dog during free play. This is worth noting, considering that studies (with healthy individuals) have shown that stroking a friendly (live) dog can have neurophysiological effects—notably, on the level of the social bonding neuropeptide oxytocin (Beetz et al., 2012)—that affect social behavior, particularly social motivation. Oxytocin has, interestingly, been implicated in the social deficits in ASD (Lerer et al., 2008; Stavropoulos & Carver, 2013), and it has been proposed that combining administration of oxytocin and behavioral interventions might diminish impaired social motivation—one problem hindering therapeutic success (Stavropoulos & Carver, 2013). Thus, it might be that during free play, adults in this study did not engage in enough contact with the dog to allow for effects on behavior to occur during the subsequent imitation task.
It is interesting that, as opposed to children, adults showed increased social contact directed at the robot than at the animal, suggesting that the robotic dog was more attractive to them than the live dog or piqued their interest in a different, more engaging manner. If mere engagement with an interesting stimulus accounted for the pattern of results observed in the child group, adults in the robot condition could have been expected to show increased social motivation and better performance during the imitation task (compared with the other two conditions). This, however, was not the case, thus pointing to particular effects among children of contact with the live dog per se.
Note that this study tested only for short-term effects. Thus, the possibility exists that the observed effects of interaction with the dog on children’s imitation behavior and social motivation may be based on novelty or attentional processes and thus may vanish over extended testing periods. Although this possibility should not be discarded, studies have reported long-term benefits associated with pet ownership in families with children with ASD (Hall et al., 2016).
Implications for Occupational Therapy Practice
This study has the following implications for occupational therapy practice:
Occupational therapy practitioners working with people with ASD may benefit from a closer look at evidence from the field of human–animal interactions.
Trained dogs used as therapeutic tools may tap the motivational states of people with ASD and help occupational therapy practitioners attain imitation goals.
Conclusion
Previous research in the particular context of occupational therapy has found evidence that “the therapeutic use of animals may be an effective way to engage a wide variety of therapy clients, as well as to enhance the effectiveness of established occupational therapy techniques” (Sams et al., 2006, p. 273). By focusing on one such established occupational therapy technique, imitation, this study adds to the field and is relevant to the profession. More important, the results reported here are preliminary; they do not indicate the utility of integrating live dogs into interventions aimed at promoting social motivation and enhancing imitation skills among people with ASD. However, this study suggests that doing so holds promise. Replication of this study on a larger scale is important so that detailed suggestions can be made for practice. Clearly, finding strategies to enhance imitation skills in people with ASD is of major clinical relevance, and research should continue to explore the potential of incorporating animals into practice.
Footnotes
Acknowledgments
Karine Silva and Mariely Lima contributed equally to this work. The authors are grateful to the participants and their families. They also gratefully acknowledge the cooperation of Associação Portuguesa para as Perturbações do Desenvolvimento e Autismo–Norte and Associación Latino-Americana de Integração in recruiting the participants and that of Concentra–Produtos Para Crianças, S.A. in providing the robotic dog. The collaboration of Ladra Comigo is also highly appreciated. Fundação para a Ciência e a Tecnologia funded Karine Silva’s participation (FCT-SFRH/BPD/100556/2014). All authors declare that they have no conflict of interest.